A new constrained blind estimation algorithm for determination of the arterial input function without reference tissues

Introduction: A major challenge in dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is the need to provide a measured estimate of the arterial input function (AIF) in order to perform pharmacokinetic modeling analysis of the resulting data. Measurement of the AIF requires the presence of a major artery in the imaging field-of-view, a significantly faster sampling rate than for tissue curve measurement, and a large dynamic range for accurate first-pass bolus quantification. AIF measurements can be confounded by signal saturation effects, partial-volume effects, and inflow and pulsatility artifacts, to name a few issues. In addition, in the brain it is known that there are local variations in AIF, highlighting the fact that the AIF in a nearby major artery may not accurately reflect the true local blood flow to tissues (1). For these reasons, development of a method of blind estimation of the AIF directly from measured tissue time curves is highly attractive (2). Here, we describe a novel, efficient, and practical approach for blind AIF estimation without use of reference tissues that is capable of recovering realistic AIFs from tissue curves alone in both simulated and real data with signal-to-noise ratios (SNR) typical for DCE-MRI data in vivo. Methods: Synthetic tissue concentration-time curves were generated by selecting values of the pharmacokinetic parameters from uniform random distributions so that , , and and Monte Carlo simulations performed to verify performance of the blind algorithm. A model functional form for the AIF based on population data from (3) was used (see (4) for details):